PRISM: Platform for Remote Sensing using Smart phones {Tathagata Das, Venkata N. Padmanabhan, Ramachandran Ramjee, Asankhaya Sharma } - Microsoft Research India. Prashanth Mohan- University of California, Berkeley

Overview Intro. Prism design. Software Sandbox. Implementation. Evaluation of few apps. Conclusion.

How it works ? Users Participate and contribute install prism runtime on their smart phone and register with the prism infrastructure. These phones run community sensing apps. App Server: Submits jobs to PRISM server. PRISM Server: Accepts jobs and deploy over registered mobile phones. Deployment: Apps use push or pull based model. -Push supports fast response, Efficiency, Scalability. PRISM Client: Registers and supports execution of jobs.

PRISM In a glance Balances interconnected goals of generality, security and scalability which is a challenge. Apps as executable binaries and Auto push. Supports efficiency and flexibility of reusing existing codes or modules. Push model ensures timely and scalable app ensuring good degree of privacy, controlled sensor info, safety against un-trusted apps.

Client environment API’s designed to enable the application server to accurately and quickly identify the set of mobile phones that can run the application. Accurate identification is enabled by a two-level predicate mechanism while quick deployment is enabled by a choice of deployment modes. Regulating Access to Sensors- No sensor, Only GPS, All sensors. An alternative is to place severe resource limits on PRISM applications that access sensitive sensors, instead of blocking such accesses entirely. Diminish privacy risks while providing greater flexibility compared to blocking access to sensors.

More on Prism. Sandbox environment with elements like -Resources metering -Taint and controlled access -Forced amnesia. Supports techniques such as system call interposition, virtual machine monitors, or capability-based systems.

Evaluation: Citizens Journalist PRISM to deliver the sensing task to a certain number of camera-equipped phones in the vicinity of the desired location. The location is specified by (lat, long) and includes a coarse radius for deployment and a fine radius for actual execution. If matching phones are not readily available, PRISM’s trigger mechanism is used to deploy at the location as and when PRISM clients register/send updates from the desired location.

Results

Party Thermometer: Human query app, Deploy based on location and sound sensor. Avoiding such unnecessary sensing would be essential for efficiency Conditions are that user to be stationary and involve if he is at the party. Music is done by performing FFT of audio samples and examining the spikes in frequency.

Road Bump Monitoring. Figure 6 shows the results. We find that of the 9 bumps detected by the application, 6 match bumps in the ground truth set within 12m (Figures 6(a) and (b))

Conclusion A need for such middleware Support flexibility, easing development and their deployment. Challenges can be addressed through infrastructural component and s\w sandbox. Balances 3 inter connected goals. Trading off a little privacy (i.e., allowing phones to be tracked within a registration interval), the push-based architecture of PRISM is able to achieve significant improvements in scalability

Q & A Thank you.